Amazon Bedrock, Azure AI Foundry, Vertex AI diverge on price

The three platforms started from the same pitch — pick a model, call an endpoint, pay per token — but the catalog counts show how differently they now count. Bedrock reports more than 100 models from 18-plus providers, Azure AI Foundry cites more than 10,000 with roughly 50 new additions a month, and Vertex AI's Model Garden sits past 200, with Azure's figure including many narrow open-weight variants and fine-tuned derivatives rather than just frontier releases.

The benchmark trackers tell a similarly split story. Anthropic's Claude Opus 5 tops SWE-bench Verified at 97.00%, yet Bedrock's own model cards still list Claude Opus 4.6 as the flagship Opus-tier offering, so Bedrock customers may be running a generation behind Anthropic's top-scoring release depending on the model ID they deploy. GPT-6 Astra leads on agent-oriented measures like Terminal-Bench v4.0, while Gemini 3.8 Flash is built for volume rather than the top of the leaderboard.

For a buyer, the decision hinges less on an abstract winner than on which constraint binds first. The headline price gap can be outweighed by Azure's regional multipliers, where an EU customer running GPT-6 Astra in the EU Data Zone now pays roughly 20% above global list price on top of a 9% increase applied earlier in the year. Capacity confirmation for launch-week access to models like GPT-6 Astra or Gemini 3.1 Pro looks likely to matter as much as list pricing.